AI Agents consume Automation AI Credits, which are part of the broader Automation credits system. These credits provide a flexible pool that can be used across different AI-powered services in Automation Anywhere.
- Automation AI Credits
-
Consumed when AI Agents run successfully. Each completed execution deducts credits from your purchased pool. These credits are metered based on token consumption per execution. See Token-based consumption metering given below for details on token ceilings.
- License requirements
-
- An Enterprise platform license includes a base entitlement of Automation AI credits. These credits are available to run AI Agents within the Public workspace.
- Private workspace:
- BYOL (Bring Your Own License) models → no Automation AI Credits are consumed for AI Agent execution.
- Automation Anywhere provided models → no Automation AI Credits are consumed for AI Agent execution, but credits will be consumed for each call to Automation Anywhere provided models.
- Public workspace:
- All executions (BYOL or Automation Anywhere provided models) consume Automation AI Credits. Additionally, each call to an Automation Anywhere provided model will consume Automation AI credits as described in Using Automation credits.
- Included credits for Enterprise and APA
-
The Enterprise or APA license credits includes a specific pool of Automation AI credits. With it you can do the following:
- No-Cost Development: You can create and test AI Agents in your private workspace without consuming credits when using your own LLM licenses (BYOL).
- Ready for Production: Your included credit pool is available immediately to run AI Agents in the public workspace. When you are ready to deploy your agentic workflow, simply check it into the public repository to begin using your existing credits.
Note: Testing AI Agents in the private workspace is free of Automation AI credit charges only when using BYOL (Bring Your Own License) models. Using Automation Anywhere provided (out-of-box) models consumes Automation AI credits in both private and public workspaces. Additionally, all executions incur third-party LLM provider costs (such as tokens) based on the model selected. - Grace policy
-
If your purchased credits are exhausted, the system allows consumption of up to 10% additional credits as a grace volume. This ensures that critical operations continue without interruption while you add or renew credits.
- Visibility
-
In the Licenses page of the Control Room, admins can view:
-
Total purchased credits
-
Used credits
-
Remaining credits available
-
Grace volume usage
-
- Token-based consumption metering
-
Each AI Agent execution is metered based on the total number of GenAI tokens (input + output) consumed. A single AI Agent execution includes up to 50,000 tokens. When an execution exceeds this ceiling, additional AI Agent units are charged. For example:
Total tokens consumed AI Agent units charged 32,500 1 52,450 2 128,900 3 Note: Token ceilings apply only when the AI Agent uses an Automation Anywhere provided model and is executed outside the AI Agent editor. Token counts include only the input and output tokens consumed across all planner invocations; tokens consumed by tools called by the agent such as AI Skills, API Tasks, or other automations are excluded. Agents built with BYOL models are not subject to token ceiling logic. - Execution consumption rules
-
Charged (credits deducted) Not charged (no deduction) Execution of a public AI Agent Execution from the Test UI Process or another Agent calls a public AI Agent Execution of a private AI Agent Execution cancelled by the system (LLM recommendation) Execution that ends in failure/error or is user-cancelled
For more details about how Automation credits work across services, see Using Automation credits.